Trending on GitHub

Graphify is an AI coding assistant skill that ingests project artifacts like code, SQL schemas, and infrastructure docs to build a unified knowledge graph. It integrates with tools such as Claude Code, Cursor, and Gemini CLI, allowing users to type /graphify to map and query their entire project context. This approach consolidates application logic, database structures, and operational documentation into a single searchable structure.

• Supports major AI coding tools including Claude Code, Cursor, and Gemini CLI.
• Ingests diverse inputs: code, SQL schemas, R scripts, shell scripts, and docs.
• Unifies app code, database schema, and infrastructure into one graph.
• Enables natural language querying of the entire project structure.

Nutlope/hallmark is a design skill for Claude Code, Cursor, and Codex that enforces original UI generation by rejecting standard LLM defaults. It selects a macrostructure, applies one of twenty themes, and runs fifty-seven anti-pattern gates before emitting code. The tool includes a 'hallmark audit' command to score existing code against these anti-patterns without making edits.

• Integrates with Claude Code, Cursor, and Codex to enforce original design patterns.
• Uses 57 'slop-test' gates and self-critique to avoid generic AI-generated looks.
• Provides a 'hallmark audit' command to score existing code against anti-patterns.
• Supports 20 distinct themes and four core verbs for structured UI generation.

This GitHub repository curates over 100 ready-to-run templates for building LLM applications, focusing on AI agents, RAG pipelines, and voice integrations. It supports major providers like Claude, Gemini, OpenAI, and Llama, offering starter code that engineers can fork and customize. The collection aims to eliminate the need to rebuild common architectural patterns from scratch for new projects.

• Provides forkable starter code for complex patterns like multi-agent teams and MCP integrations
• Covers diverse modalities including voice agents, fine-tuning workflows, and always-on agents
• Supports a broad vendor stack including OpenAI, xAI, Qwen, and Llama out of the box
• Includes step-by-step tutorials to accelerate development and reduce initial setup friction

Andrew Kelley, creator of Zig, criticized a recent port of his language to Bun, which was executed by an AI using Claude and Rust. He labeled the resulting code as 'unreviewed slop,' highlighting the speed and low cost of the 11-day, $165,000 API-driven effort. The comment underscores skepticism around AI-generated code quality when human review is bypassed.

• AI-driven ports can be delivered in days but may lack rigorous code review.
• High API costs ($165k) do not guarantee engineering quality or maintainability.
• Language creators may reject AI-generated ports that skip human oversight.
• Rust rewrites via LLMs require strict validation before integration.

OpenCut is an open-source video editor being rewritten from scratch using a Rust core to support web, desktop, and mobile platforms from a single codebase. The new architecture prioritizes an Editor API and first-class third-party plugins, enabling a plugin-first design. Key upcoming features include an MCP server for AI agent integration, headless mode for automation, and a built-in scripting tab.

• Rust core enables cross-platform consistency across web, desktop, and mobile.
• MCP server support allows direct integration with AI agents for automation.
• Plugin-first architecture and Editor API facilitate robust third-party extensions.
• Headless mode and scripting tab support batch rendering and CI/CD workflows.
• Classic version remains available; rewrite is currently in beta at new.opencut.app.